Why my forecast with Croston methods are always constant?

Question

I'm using the library tsintermittent to forecast the demand of a slow-moving demand item. I'm using 36 monthly observations and you can see them in the code.

I want to forecast the next 10 months after my observations using the function crost(), so I did the following code:

prueba5_1 <- c(0,0,0,0,0,0,0,0,0,0,0,5,5480,0,0,0,1520,3001,1001,3000,3000,2000,2000,0,0,1000,1000,1000,0,0,0,1,1000,0,3000,0)
prueba5_1_ts <- ts(prueba5_1, start = c(2017,12), end = c(2020,11), frequency = 12)
prueba5_1_ts_crost <- crost(prueba5_1_ts, h = 10, outplot = 1)$frc.out

I already know that when I use the $frc.out I can get the output of my prediction and that's whys is very strange to me that the next 10 months are showing the same values as follows:

[1] 808.3602 808.3602 808.3602 808.3602 808.3602 808.3602 808.3602 808.3602 808.3602 808.3602

I'm new trying to understand this methods, but it would be great if someone can guide me in understand if all the predicted periods are the same or I'm doing something wrong with my code.

Thank you in advance.

Answer 1

$frc.out estimate the forecast for average demand. It means that this is a constant because the Croston method is not able to forecast the trend or seasonality. This post explains extensively what the crost function does. https://stats.stackexchange.com/questions/127337/explain-the-croston-method-of-r